Call large indel with small genomes

[ChenddatHKU]

Hi Sir/Madam,

Thank you all for making this amazing caller!

Is there anyone trying to apply cuteSV for virus SV calling?

We focus on the influenza virus and large internal deletions; its genome is relatively small (13 kb, 8 chromosomes).
Typically, sequencing depth is 200x (could reach 10^3-5 depth). However, the overall frequency of reads carrying deletion fluctuates from ~1% to 99% (similar to somatic mutation?), and individual SVs exhibit extensive breakpoint heterogeneity (e.g., >100 deletion types per chromosome, with their breakpoints clustered around a ~100-200 bp region) (heterozygous SV). A schematic graph of breakpoints detected with NGS is shown below (PMID: 38407436).

Considering the high error rate of ONT reads, do you think it's possible to call the breakpoints confidently with cuteSV?
Given this complexity, I suspect adjustments may be needed, e.g, not only for minimum read support thresholds. Do you have recommended settings for such cases?

Thank you so much for your suggestions.

Best regards

[Meltpinkg]

Hello, @ChenddatHKU

To be honest, I have not previously applied cuteSV to virus genome analysis.

Since the variant breakpoint extraction does not vary much through different species, it's possible to call the breakpoints of virus variants. However, since the genotype module in cuteSV is designed for diploid genomes, it may be totally suboptimal for these somatic mutations.

For this instance, I have a few suggestions: --min_support 4 (in fact, this low threshold would retain the SVs of low frequency, with the sacrifies of more potential false positives; therefore, several filters may further need); --max_cluster_bias_DEL 100 --diff_ratio_merging_DEL 0.3 (this lower threshold works better for ONT reads of high error rate).

The outlined suggestions are based on my practical experience. If you encounter any issues during implementation, feel free to reach out. I’ll be glad to assist and follow up as needed.

Best regards,
Shuqi